https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65482
--- Comment #4 from Bill Schmidt <wschmidt at gcc dot gnu.org> --- OK, that explains how we got here. At the moment, the usage of the flag only matters to the test suite when running on older hardware. On P8, the test suite uses -mpower8-vector rather than -mvsx -mno-allow-movmisalign. So in that sense this will "age out" over time, and we are getting good vectorization coverage on P8, so it is probably best just to leave things as they are. We should clean up tests that aren't working properly (like the one from PR65484), though. Martin, not sure where you'd like to go with this bug. The flag is undocumented on purpose, and from Mike's recollections this was purposely used in the test suite to avoid unprofitable vectorization on the older hardware. I'd be inclined to suggest that we close the bug, but I'll leave that decision up to you. Thanks!
